全能加载图像🐠meeeyo.com

ComfyUI's built-in LoadImage is strictly local - it reads from your input directory, full stop. That's a wall the moment your image lives on a URL, in another folder, or as a base64 string inside a workflow's data. GenericImageLoader (全能加载图像 - "omnipotent image loader") is that wall coming down: one STRING input where you type or wire a path, an http(s) URL, or a base64 blob, and you get a standard IMAGE out. It's the loader this pack uses internally everywhere, and it earns the "generic" in its name.

How it works

The single image_input string is auto-detected in three ways, in this order:

• Starts with http:// or https:// → it downloads the image (the pack uses requests, so that dependency needs to be present).
• Is a file on disk (os.path.isfile) → opens it directly. This means it reads absolute paths anywhere on your machine, not just the input folder - which is both the point and the security footnote below.
• Anything else → treated as base64. It handles both the data:image/png;base64,... prefixed form and the bare base64 string.

It also strips Unicode direction-control characters from the start of paths - a nice touch for Windows users who copy paths that come in with invisible bidi characters. Outputs are two: an IMAGE tensor and a MASK. The mask is derived from the alpha channel - pixels with alpha > 0 become mask-white. So a transparent PNG gives you a usable mask for compositing or inpainting without a separate step. (Note: it converts to RGBA first, so the IMAGE output is 4-channel RGBA, which some downstream nodes will want composited to RGB.)

Why you'd reach for it

Three workflows, one node: fetching a reference image from a URL (no download step first), batch-loading images from an arbitrary folder path that LoadImage can't see, and reading images that arrive as base64 - which is how a lot of API and web-tool outputs hand images around. It composes with ImageAdjuster from the same pack (resize/snap the result) and ImageOverlayAlignment (stack layers) to build a full image-prep pipeline.

Install

Part of ComfyUI_StringOps:

cd ComfyUI/custom_nodes
git clone https://github.com/MeeeyoAI/ComfyUI_StringOps.git

Restart. Or ComfyUI Manager → "ComfyUI_StringOps". Requires requests for the URL path - it's in the pack's requirements.txt, and Manager installs it. No model files. Chinese UI (全能加载图像) with the 🐠meeeyo.com branding and a WeChat ad in the description.

Common issues

Three real ones. Error handling is thin: a bad URL or corrupt base64 raises an exception and the node fails - there's no graceful fallback, so validate your input before you queue. Absolute paths = a footgun: because it reads any path, a workflow that worked on someone else's machine will reference their paths; shared workflows with this node need path awareness. And keep in mind the file check runs after the URL check but before base64 - a string that accidentally resolves to a file path on your system will load that file instead of being decoded, which can be confusing if you're switching between base64 and local modes. It's the one loader you'll keep handy once you start pulling images from the internet.